Cricket Predictions

Data Collection

We proceeded the process by collecting the relevant data for the total player for both teams. So in order to capture the present form of all the players from both team( India and Australia) we took in account of their recent matches. To capture the form against the opponent we took in account all the matches played by the current teams against one another. We also consider how each ground plays a role in determining each players favorite ground to play. The data we had covered various things. Batsman - Total ball faced by the batsman, total run scored, Strike rate, number of 4s and 6s, The position the batsman came to bat, the result of coin toss, Bowler - The total number of overs bowled by the bowler, number of maiden overs, total wickets taken, economy, the total run given,

Data Cleaning

Removed whitespaces before and after the entries, Subsituted a common value ‘NA’ for all kinds of missing entries like {blanks}, {-}, {–}, {null}, Correction of the spellings of the names of the batsman and bowlers

Importing the Data for the Batsman in from the excel file. Removing unnecessory columns like ‘Date’

library(readxl)
Batsman <- read_excel("C:/Users/Arjit/Downloads/Batsman.xlsx", sheet = "Batsman", col_types = c("text", "numeric", "text", "numeric", "numeric", "numeric", "numeric", "numeric", "text", "numeric", "text", "text", "date"))
Batsman <- subset(Batsman, select = -`Start Date` )

Changing the categorical Variables to factors before building the model

Batsman$Batsman<-factor(Batsman$Batsman)
Batsman$Pos<-factor(Batsman$Pos)
Batsman$Dismissal<-factor(Batsman$Dismissal)
Batsman$Inns<-factor(Batsman$Inns)
Batsman$Opposition<-factor(Batsman$Opposition)
Batsman$Ground<-factor(Batsman$Ground)

Subsituting NAs for the feilds marked as ‘-’

Batsman$`4s`[Batsman$`4s`=="-"]<-NA
Batsman$`6s`[Batsman$`6s`=="-"]<-NA
Batsman$SR[Batsman$SR=="-"]<-NA
Batsman$Pos[Batsman$Pos=="-"]<-NA
Batsman$Dismissal[Batsman$Dismissal=="-"]<-NA
Batsman$Inns[Batsman$Inns=="-"]<-NA
Batsman$Opposition[Batsman$Opposition=="-"]<-NA
Batsman$Ground[Batsman$Ground=="-"]<-NA
Batsman$Batsman[Batsman$Batsman=="-"]<-NA
Batsman$Mins[Batsman$Mins=="-"]<-NA
Batsman$BF[Batsman$BF=="-"]<-NA

Changing discrete variables to numeric to tell are they are a number

Batsman$Mins<-as.numeric(Batsman$Mins)
Batsman$Runs<-as.numeric(Batsman$Runs)
Batsman$BF<-as.numeric(Batsman$BF)
Batsman$SR<-as.numeric(Batsman$SR)
Batsman$`4s`<-as.numeric(Batsman$`4s`)
Batsman$`6s`<-as.numeric(Batsman$`6s`)

Creating Visualizations to Understand the data.

We will plot the batsman with the number of matches they have played. This will help us in knowing which players are more experienced.

Batsman_Experience<-as.data.frame(table(Batsman$Batsman))

library(plotly)

## Loading required package: ggplot2

## 
## Attaching package: 'plotly'

## The following object is masked from 'package:ggplot2':
## 
##     last_plot

## The following object is masked from 'package:stats':
## 
##     filter

## The following object is masked from 'package:graphics':
## 
##     layout

plot_ly(
  x = Batsman_Experience$Var1,y = Batsman_Experience$Freq,name = "Batsman_Experience",type = "bar")

Here, we are trying to see a plot of the batsman vs number of Runs scored by him.

batsman_runs <-data.frame(Batsman_Name = c("Rohit Sharma","Virat Kohli","Shikhar Dhawan", "Ravindra Jadeja","KL Rahul", "Ambati Rayudu","Aaron Finch","Shaun Marsh","Glenn Maxwell","Usman Khawaja","Marcus Stoinis"),
Batsman_Runs = c(sum(Batsman$Runs[Batsman$Batsman =="Rohit Sharma"]),sum(Batsman$Runs[Batsman$Batsman =="Virat Kohli"]),sum(Batsman$Runs[Batsman$Batsman =="Shikhar Dhawan"]),sum(Batsman$Runs[Batsman$Batsman =="Ravindra Jadeja"]),
              sum(Batsman$Runs[Batsman$Batsman =="KL Rahul"]),sum(Batsman$Runs[Batsman$Batsman =="Ambati Rayudu"]),
              sum(Batsman$Runs[Batsman$Batsman =="Aaron Finch"]),sum(Batsman$Runs[Batsman$Batsman =="Shaun Marsh"]),
              sum(Batsman$Runs[Batsman$Batsman =="Glenn Maxwell"]),sum(Batsman$Runs[Batsman$Batsman =="Usman Khawaja"]),sum(Batsman$Runs[Batsman$Batsman =="Marcus Stoinis"])))  

batsman_runs

library(plotly)

plot_ly(
  x = batsman_runs$Batsman_Name,y = batsman_runs$Batsman_Runs,name = "Batsman Runs",type = "bar")

Our plot shows us that Rohit Sharma has the highest number of Runs, followed by Virat Kohli

Similarily, we will plot for the number of fours

batsman_fours=data.frame(Batsman_Name = c("Rohit Sharma","Virat Kohli","Shikhar Dhawan", "Ravindra Jadeja","KL Rahul", "Ambati Rayudu","Aaron Finch","Shaun Marsh","Glenn Maxwell","Usman Khawaja","Marcus Stoinis"),
Batsman_fours = c(sum(Batsman$`4s`[Batsman$Batsman =="Rohit Sharma"]),sum(Batsman$`4s`[Batsman$Batsman =="Virat Kohli"]),sum(Batsman$`4s`[Batsman$Batsman =="Shikhar Dhawan"]),sum(Batsman$`4s`[Batsman$Batsman =="Ravindra Jadeja"]),
                                           sum(Batsman$`4s`[Batsman$Batsman =="KL Rahul"]),sum(Batsman$`4s`[Batsman$Batsman =="Ambati Rayudu"]),
                                           sum(Batsman$`4s`[Batsman$Batsman =="Aaron Finch"]),sum(Batsman$`4s`[Batsman$Batsman =="Shaun Marsh"]),
                                           sum(Batsman$`4s`[Batsman$Batsman =="Glenn Maxwell"]),sum(Batsman$`4s`[Batsman$Batsman =="Usman Khawaja"]),sum(Batsman$`4s`[Batsman$Batsman =="Marcus Stoinis"])))                          

batsman_fours

plot_ly(
  x = batsman_fours$Batsman_Name,y = batsman_fours$Batsman_fours,name = "Batsman_fours",type = "bar")

Similarily, for sixes

library(plotly)


batsman_sixes <-data.frame(Batsman_Name = c("Rohit Sharma","Virat Kohli", "Shikhar Dhawan", "Ravindra Jadeja","KL Rahul", "Ambati Rayudu","Aaron Finch","Shaun Marsh","Glenn Maxwell","Usman Khawaja","Marcus Stoinis"),
                          Batsman_Sixes = c(sum(Batsman$`6s`[Batsman$Batsman =="Rohit Sharma"]),sum(Batsman$`6s`[Batsman$Batsman =="Virat Kohli"]),sum(Batsman$`6s`[Batsman$Batsman =="Shikhar Dhawan"]),sum(Batsman$`6s`[Batsman$Batsman =="Ravindra Jadeja"]),
                                           sum(Batsman$`6s`[Batsman$Batsman =="KL Rahul"]),sum(Batsman$`6s`[Batsman$Batsman =="Ambati Rayudu"]),
                                           sum(Batsman$`6s`[Batsman$Batsman =="Aaron Finch"]),sum(Batsman$`6s`[Batsman$Batsman =="Shaun Marsh"]),
                                           sum(Batsman$`6s`[Batsman$Batsman =="Glenn Maxwell"]),sum(Batsman$`6s`[Batsman$Batsman =="Usman Khawaja"]),sum(Batsman$`6s`[Batsman$Batsman =="Marcus Stoinis"])))                         

plot_ly(
  x = batsman_sixes$Batsman_Name,y = batsman_sixes$Batsman_Sixes,name = "Batsman Runs",type = "bar")

Creating Dummy Variables of Categorical Function to use in Regression

Batsman<-data.frame(model.matrix(~ .,Batsman))

Splitting the data int Testing and Training and applying split as 0.75.

require(caTools) 

## Loading required package: caTools

set.seed(123)
sample = sample.split(Batsman,SplitRatio = 0.75)
Batsman_train =subset(Batsman,sample ==TRUE)
Batsman_test=subset(Batsman, sample==FALSE)

Building a simple linear regression model to predict the runs

Batsman_model<-lm(Runs~.,data=Batsman_train)


k<-as.data.frame(subset(Batsman_test, select = - Runs ))
y<-predict.lm(Batsman_model,k)

mean(y-Batsman_test$Runs) *mean(y-Batsman_test$Runs) #Mean Squared Error as a metric to evaluate the model

## [1] 0.1653363

We will now try building a random forest regressor model. This will also help us decide the important features for variable sleection

require(randomForest)

## Loading required package: randomForest

## randomForest 4.6-14

## Type rfNews() to see new features/changes/bug fixes.

## 
## Attaching package: 'randomForest'

## The following object is masked from 'package:ggplot2':
## 
##     margin

fit=randomForest(Runs~.,data=Batsman_train,importance=T)
y2<-predict(fit,k)
mean(Batsman_test$Runs-y2)*mean(Batsman_test$Runs-y2)

## [1] 0.03424071

We can see that the mean sqaured error has reduced a lot upon using random forest for building the model.

Now, we will look at the important features

library(caret)

## Loading required package: lattice

importance <- varImp(fit, scale=FALSE)
varImpPlot(fit)

Till now we have studied the data using the historic data. Now, we build our own data according to the upcoming match details.This data is stored in parameter_Batsman.

Reading the data and converting the datatypes.

Creating the dummy variables.

library(readxl)

parameter_Batsman <- read_excel("C:/Users/Arjit/Downloads/parameter_Batsman (1).xlsx",sheet = "Batsman", col_types = c("text", "text", "text", "text", "text"))
View(parameter_Batsman)

parameter_Batsman$Ground<-factor(parameter_Batsman$Ground)
parameter_Batsman$Opposition<-factor(parameter_Batsman$Opposition)
parameter_Batsman$Batsman<-factor(parameter_Batsman$Batsman)
parameter_Batsman$Pos<-factor(parameter_Batsman$Pos)
parameter_Batsman$Inns<-factor(parameter_Batsman$Inns)

parameter_Batsman<-data.frame(model.matrix(~ .,parameter_Batsman))

Fitting a random forest regressor model to the data and predicting who will score the highest number of Runs.

set.seed(123) #To produce reproducible results

fit_final<-randomForest(Runs~GroundHyderabad..Deccan. + GroundMohali +GroundNagpur +GroundRanchi+Oppositionv.India+BatsmanAmbati.Rayudu+BatsmanGlenn.Maxwell + BatsmanKL.Rahul+BatsmanMarcus.Stoinis+BatsmanRavindra.Jadeja+ BatsmanRohit.Sharma+BatsmanShaun.Marsh+ BatsmanUsman.Khawaja+BatsmanVirat.Kohli +Pos2+ Pos3+ Pos4+ Pos5 +Pos6 +  Inns2,data = Batsman)

##fitted a random forest model 
k_final<-parameter_Batsman[c("GroundHyderabad..Deccan.","GroundMohali","GroundNagpur","GroundRanchi","Oppositionv.India","BatsmanAmbati.Rayudu","BatsmanGlenn.Maxwell","BatsmanKL.Rahul","BatsmanMarcus.Stoinis","BatsmanRavindra.Jadeja","BatsmanRohit.Sharma","BatsmanShaun.Marsh","BatsmanUsman.Khawaja","BatsmanVirat.Kohli","Pos2","Pos3","Pos4","Pos5","Pos6","Inns2")] ##New test data to provide to the model to predict results

parameter_Batsman$y_final=predict(fit_final,k_final) #Predicting Results
parameter_Batsman$y_final==max(parameter_Batsman$y_final) #Selecting the parameters with predicted highest number of runs

##  [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [12] FALSE FALSE FALSE FALSE  TRUE FALSE FALSE FALSE FALSE FALSE FALSE
## [23] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [34] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [45] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE

Our model shows the value as TRUE for 16th Row. This means that the new data we gave (According to the upcoming match) has been predicted by our model and the 16th Row has the highest number of Runs.

We can see the corresponding value of ‘Batsman Name’ to find out who is scoring the maximum runs

(parameter_Batsman[16,]==1)

##    X.Intercept. GroundHyderabad..Deccan. GroundMohali GroundNagpur
## 16         TRUE                    FALSE        FALSE         TRUE
##    GroundRanchi Oppositionv.India BatsmanAmbati.Rayudu
## 16        FALSE             FALSE                FALSE
##    BatsmanGlenn.Maxwell BatsmanKL.Rahul BatsmanMarcus.Stoinis
## 16                FALSE           FALSE                 FALSE
##    BatsmanRavindra.Jadeja BatsmanRohit.Sharma BatsmanShaun.Marsh
## 16                  FALSE                TRUE              FALSE
##    BatsmanUsman.Khawaja BatsmanVirat.Kohli  Pos2  Pos3  Pos4 Pos5  Pos6
## 16                FALSE              FALSE FALSE FALSE FALSE TRUE FALSE
##    Inns2 y_final
## 16 FALSE   FALSE

As it is Showing True only infront of Rohit Sharma, according to our model, Rohit Sharma will have the maximum runs = 79.78106 or 78 as shown in the next chunk

max(parameter_Batsman$y_final) #maximum value of runs 

## [1] 79.78106

Similarily, For predicting the highest number of fours

fit_final_1<-randomForest(X.4s.~GroundHyderabad..Deccan. + GroundMohali +GroundNagpur +GroundRanchi+Oppositionv.India+BatsmanAmbati.Rayudu+BatsmanGlenn.Maxwell + BatsmanKL.Rahul+BatsmanMarcus.Stoinis+BatsmanRavindra.Jadeja+ BatsmanRohit.Sharma+BatsmanShaun.Marsh+ BatsmanUsman.Khawaja+BatsmanVirat.Kohli +Pos2+ Pos3+ Pos4+ Pos5 +Pos6 +  Inns2,data = Batsman)

#fitted the model


k_final<-parameter_Batsman[c("GroundHyderabad..Deccan.","GroundMohali","GroundNagpur","GroundRanchi","Oppositionv.India","BatsmanAmbati.Rayudu","BatsmanGlenn.Maxwell","BatsmanKL.Rahul","BatsmanMarcus.Stoinis","BatsmanRavindra.Jadeja","BatsmanRohit.Sharma","BatsmanShaun.Marsh","BatsmanUsman.Khawaja","BatsmanVirat.Kohli","Pos2","Pos3","Pos4","Pos5","Pos6","Inns2")]
#new data to be given for predictions
parameter_Batsman$y_final_1=predict(fit_final_1,k_final) #Predicting for fours
parameter_Batsman$y_final_1==max(parameter_Batsman$y_final_1) #Selecting row with highest number of fours predicted

##  [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [12] FALSE FALSE  TRUE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [23] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [34] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [45] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE

It is showing the true value for row number 14

(parameter_Batsman[14,]==1) #looking at the parametersin row 14

##    X.Intercept. GroundHyderabad..Deccan. GroundMohali GroundNagpur
## 14         TRUE                    FALSE        FALSE         TRUE
##    GroundRanchi Oppositionv.India BatsmanAmbati.Rayudu
## 14        FALSE             FALSE                FALSE
##    BatsmanGlenn.Maxwell BatsmanKL.Rahul BatsmanMarcus.Stoinis
## 14                FALSE           FALSE                 FALSE
##    BatsmanRavindra.Jadeja BatsmanRohit.Sharma BatsmanShaun.Marsh
## 14                  FALSE               FALSE              FALSE
##    BatsmanUsman.Khawaja BatsmanVirat.Kohli  Pos2 Pos3  Pos4  Pos5  Pos6
## 14                FALSE               TRUE FALSE TRUE FALSE FALSE FALSE
##    Inns2 y_final y_final_1
## 14 FALSE   FALSE     FALSE

As it is Showing True only infront of Virat Kohli, according to our model, Virat Kohli will have the maximum fours = 6.432789 or 6 as shown in the next chunk

max(parameter_Batsman$y_final_1) #Maximum number of fours

## [1] 6.432789

For predicting the highest number of sixes

fit_final_2<-randomForest(X.6s.~GroundHyderabad..Deccan. + GroundMohali +GroundNagpur +GroundRanchi+Oppositionv.India+BatsmanAmbati.Rayudu+BatsmanGlenn.Maxwell + BatsmanKL.Rahul+BatsmanMarcus.Stoinis+BatsmanRavindra.Jadeja+ BatsmanRohit.Sharma+BatsmanShaun.Marsh+ BatsmanUsman.Khawaja+BatsmanVirat.Kohli +Pos2+ Pos3+ Pos4+ Pos5 +Pos6 +  Inns2,data = Batsman)

#fitted the model
k_final<-parameter_Batsman[c("GroundHyderabad..Deccan.","GroundMohali","GroundNagpur","GroundRanchi","Oppositionv.India","BatsmanAmbati.Rayudu","BatsmanGlenn.Maxwell","BatsmanKL.Rahul","BatsmanMarcus.Stoinis","BatsmanRavindra.Jadeja","BatsmanRohit.Sharma","BatsmanShaun.Marsh","BatsmanUsman.Khawaja","BatsmanVirat.Kohli","Pos2","Pos3","Pos4","Pos5","Pos6","Inns2")]

#new data
parameter_Batsman$y_final_2=predict(fit_final_2,k_final) #predcitions for number of sixes
parameter_Batsman$y_final_2==max(parameter_Batsman$y_final_2) #fetching row with highest number of sixes

##  [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [12] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [23] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [34]  TRUE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [45] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE

It is showing the true value for row number 34

(parameter_Batsman[34,]==1) #looking at parameters of row 34

##    X.Intercept. GroundHyderabad..Deccan. GroundMohali GroundNagpur
## 34         TRUE                    FALSE         TRUE        FALSE
##    GroundRanchi Oppositionv.India BatsmanAmbati.Rayudu
## 34        FALSE             FALSE                FALSE
##    BatsmanGlenn.Maxwell BatsmanKL.Rahul BatsmanMarcus.Stoinis
## 34                FALSE           FALSE                 FALSE
##    BatsmanRavindra.Jadeja BatsmanRohit.Sharma BatsmanShaun.Marsh
## 34                  FALSE                TRUE              FALSE
##    BatsmanUsman.Khawaja BatsmanVirat.Kohli  Pos2  Pos3  Pos4  Pos5  Pos6
## 34                FALSE              FALSE FALSE FALSE FALSE FALSE FALSE
##    Inns2 y_final y_final_1 y_final_2
## 34 FALSE   FALSE     FALSE     FALSE

As it is Showing True only infront of Rohit Sharma, according to our model, Rohit Sharma will have the maximum fours = 2.707413 or 3 as shown in the next chunk

max(parameter_Batsman$y_final_2) #maximum number of sixes

## [1] 2.707413

Same Process will be followed for Bowlers data

Here again we will import the data, Change the datatypes, replace the ‘NA’ values

library(readxl)
Bowler <- read_excel("C:/Users/Arjit/Downloads/Batsman.xlsx", sheet = "Bowler", col_types = c("text", "numeric", "numeric", "numeric", "numeric", "numeric", "numeric", "numeric", "text", "text", "date"))

#Imported the data

Bowler <- subset(Bowler, select = -`Start Date` )

Bowler$Bowler<-factor(Bowler$Bowler)
Bowler$Pos<-factor(Bowler$Pos)
Bowler$Mdns<-factor(Bowler$Mdns)
Bowler$Inns<-factor(Bowler$Inns)
Bowler$Opposition<-factor(Bowler$Opposition)
Bowler$Ground<-factor(Bowler$Ground)

Bowler$Overs[Bowler$Overs=="-"]<-NA
Bowler$Runs[Bowler$Runs=="-"]<-NA
Bowler$Wkts[Bowler$Wkts=="-"]<-NA
Bowler$Pos[Bowler$Pos=="-"]<-NA
Bowler$Mdns[Bowler$Mdns=="-"]<-NA
Bowler$Inns[Bowler$Inns=="-"]<-NA
Bowler$Opposition[Bowler$Opposition=="-"]<-NA
Bowler$Ground[Bowler$Ground=="-"]<-NA
Bowler$Bowler[Bowler$Bowler=="-"]<-NA
Bowler$Econ[Bowler$Econ=="-"]<-NA

Plotting the data to see the relationship between the parameters

This plot shows us the number of matches played by any bowler. This helps us see the most experienced players.

Bowler_Experience<-as.data.frame(table(Bowler$Bowler))

library(plotly)

plot_ly(
  x = Bowler_Experience$Var1,y = Bowler_Experience$Freq,name = "Bowler Experience",type = "bar")

The following plot shows us the number of wickets by each of the bowlers

Bowler_Wickets <-data.frame(Bowler_Name = c("Nathan Lyon","Pat Cummins","Ravindra Jadeja","Kuldeep Yadav","Kane Richardson","Glenn Maxwell","Adam Zampa","Yuzvendra Chahal","Nathan Coulter-Nile","Mohammed Shami","Jasprit Bumrah","Bhuvneshwar Kumar"),
Bowler_Plus = c(sum(Bowler$Wkts[Bowler$Bowler =="Nathan Lyon"]),sum(Bowler$Wkts[Bowler$Bowler =="Pat Cummins"]),sum(Bowler$Wkts[Bowler$Bowler 
              =="Ravindra Jadeja"]),sum(Bowler$Wkts[Bowler$Bowler =="Kuldeep Yadav"]),sum(Bowler$Wkts[Bowler$Bowler =="Kane Richardson"]),
              sum(Bowler$Wkts[Bowler$Bowler =="Adam Zampa"]),sum(Bowler$Wkts[Bowler$Bowler =="Bhuvneshwar Kumar"]),
              sum(Bowler$Wkts[Bowler$Bowler =="Yuzvendra Chahal"]),sum(Bowler$Wkts[Bowler$Bowler =="Nathan Coulter-Nile"]),
              sum(Bowler$Wkts[Bowler$Bowler =="Glenn Maxwell"]),sum(Bowler$Wkts[Bowler$Bowler =="Jasprit Bumrah"]),sum(Bowler$Wkts[Bowler$Bowler =="Mohammed Shami"])))

library(plotly)

plot_ly(
  x = Bowler_Wickets$Bowler_Name,y = Bowler_Wickets$Bowler_Plus,name = "Bowler Wicket",type = "bar")

Changing the data types of the variables, splitting the data

#Changing the Data types
Bowler$Runs<-as.numeric(Bowler$Runs)
Bowler$Wkts<-as.numeric(Bowler$Wkts)
Bowler$Econ<-as.numeric(Bowler$Econ)
Bowler$Mdns<-as.numeric(Bowler$Mdns)
Bowler$Overs<-as.numeric(Bowler$Overs)
Bowler<-data.frame(model.matrix(~ .,Bowler))

Bowler <- subset(Bowler, select = - X.Intercept. )

#Splitting Data in the Ratio 0.8

require(caTools) 
set.seed(123)
sample = sample.split(Bowler,SplitRatio = 0.80)
Bowler_train =subset(Bowler,sample ==TRUE)
Bowler_test=subset(Bowler, sample==FALSE)

Creating a simple linear regression model and looking at the mean squared error

Bowler_model<-lm(Wkts~.,data=Bowler_train)


k<-as.data.frame(subset(Bowler_test, select = - Wkts ))
y<-predict.lm(Bowler_model,k)

mean(y-Bowler_test$Wkts)*mean(y-Bowler_test$Wkts)

## [1] 0.001599208

Mean Squared Error given by this model is 0.001599208

Now creating a random forest model and looking at the mean squared error. Our aim is to decrease the mean squared error

set.seed(123)
require(randomForest)
fit=randomForest(Wkts~.,data=Bowler_train,importance=T)
y2<-predict(fit,k)
mean(Bowler_test$Wkts-y2)*mean(Bowler_test$Wkts-y2)

## [1] 0.0006857448

Mean Squared Error given by this model is 0.0006857448 which is very much lessor than the previous model. That is why this is a better model.

The following plot will show us the important variables which will help us in feature selection.

library(caret)
importance <- varImp(fit, scale=FALSE)
varImpPlot(fit)

Our analysis was based on the historic data. Now we will use the data for the current match and build the model.

#Reading the data

library(readxl)
parameter_Bowler <- read_excel("C:/Users/Arjit/Downloads/parameter_Batsman (1).xlsx", sheet = "Bowler", col_types = c("text",  "text", "text", "text", "text"))
parameter_Bowler$Ground<-factor(parameter_Bowler$Ground)
parameter_Bowler$Opposition<-factor(parameter_Bowler$Opposition)
parameter_Bowler$Bowler<-factor(parameter_Bowler$Bowler)
parameter_Bowler$Pos<-factor(parameter_Bowler$Pos)
parameter_Bowler$Inns<-factor(parameter_Bowler$Inns)

#Creating dummy Variables

parameter_Bowler<-data.frame(model.matrix(~ .,parameter_Bowler))

set.seed(123)

#Building the model
fit_final <-randomForest(Wkts~ GroundHyderabad..Deccan. + GroundMohali+GroundNagpur  + GroundRanchi  +Oppositionv.India+BowlerGlenn.Maxwell + BowlerJasprit.Bumrah+BowlerKane.Richardson+BowlerPat.Cummins+BowlerRavindra.Jadeja+BowlerYuzvendra.Chahal+Pos2 +Pos3 +Pos4+Pos5 +Pos6 +Inns2+BowlerKuldeep.Yadav+BowlerMohammed.Shami+BowlerNathan.Lyon,data = Bowler,importance=T)

#Putting in the new data

k_final<-parameter_Bowler[c( "GroundHyderabad..Deccan.","GroundMohali" ,  "GroundNagpur" , "GroundRanchi" ,    "Oppositionv.India", "BowlerGlenn.Maxwell" , "BowlerJasprit.Bumrah", "BowlerKane.Richardson" , "BowlerKuldeep.Yadav", "BowlerMohammed.Shami", "BowlerNathan.Lyon" ,   "BowlerPat.Cummins", "BowlerRavindra.Jadeja","BowlerYuzvendra.Chahal" , "Pos2" , "Pos3" ,"Pos4" ,  "Pos5","Pos6", "Inns2")]

#Predicting data for number of wickets as dependent variable 

parameter_Bowler$y_final=predict(fit_final,k_final)

#Fetching the row with highest number of wickets
parameter_Bowler$y_final==max(parameter_Bowler$y_final)

##  [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [12] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [23] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [34] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
## [45] FALSE FALSE FALSE FALSE FALSE FALSE  TRUE FALSE FALSE FALSE FALSE

It shows ‘TRUE’ for row 51

parameter_Bowler[51,]==1 #fetching all the parameters for row 51

##    X.Intercept. GroundHyderabad..Deccan. GroundMohali GroundNagpur
## 51         TRUE                    FALSE        FALSE        FALSE
##    GroundRanchi Oppositionv.India BowlerGlenn.Maxwell BowlerJasprit.Bumrah
## 51        FALSE             FALSE               FALSE                FALSE
##    BowlerKane.Richardson BowlerKuldeep.Yadav BowlerMohammed.Shami
## 51                 FALSE                TRUE                FALSE
##    BowlerNathan.Lyon BowlerPat.Cummins BowlerRavindra.Jadeja
## 51             FALSE             FALSE                 FALSE
##    BowlerYuzvendra.Chahal Pos2  Pos3  Pos4  Pos5  Pos6 Inns2 y_final
## 51                  FALSE TRUE FALSE FALSE FALSE FALSE  TRUE   FALSE

As it is showing ‘TRUE’ for Kuldeep Yadav, Our model predicts that Kuldeep Yadav will have the highest number of wickets that is 2 wickets as shown in the next chunk.

max(parameter_Bowler$y_final) #max number of wickets

## [1] 2.216886

Predictions

Winner of the Series : INDIA

As the highest number of wickets, runs, fours, and sixes are by Indian Batsman and Bowlers,

Series Output: 4-0 (India,Australia)

Highest Run Scorer : Rohit Sharma

Highest Wicket Taker: Kuldeep Yadav

Maximum Sixes : Rohit Sharma

Maximum Fours : Virat Kohli